Quick Answer
Call tracking gets expensive when you rent a separate number for every ad, ad group, and test. Dynamic number insertion only needs enough numbers for peak concurrent visitors on the site, not one line per creative. Size the pool to concurrency, use static numbers for offline placements, and keep Smart Bidding on qualified or booked calls.
The bill blows up when you buy numbers like ad IDs
Most Google Ads accounts that depend on phone leads eventually add call tracking. The first invoice looks fine. Then someone maps a unique number to every ad, every ad group, and every landing page variant. The pool grows with the test calendar instead of with traffic. Extra numbers stack. Minute overages show up. The tracking line item starts to feel like another media buy.
That mental model is the problem. Call tracking numbers are temporary labels for a session, not permanent IDs for every creative. Treat them like ad IDs and cost scales with how many experiments you run, not with how many people are on the site at once.
I see this a lot with growing local lead gen accounts. The media plan adds Phrase match and Exact match variants, a few Call Only Ads, and a couple of landing page tests. The tracking plan tries to keep pace one number at a time. The dashboard still looks useful, but the monthly bill starts to punish the testing habit that should improve the account.
What you actually need to buy
Dynamic number insertion swaps a trackable number onto the page for each active visitor. The tool only needs a unique number for each concurrent session that might call. If twenty people can be on the site at peak with a real chance of dialing, you need a pool sized for that concurrency, not a number for every RSA in the account.
Static numbers still matter. Print, truck wraps, TV, and other offline placements need fixed lines. Website call conversions from paid search do not. For the site, one dynamic pool shared across campaigns is usually enough to start. You can tighten later when the data earns it.
Minutes are another scare line on the invoice. Inbound tracked minutes are what you pay for. Outbound callbacks from sales usually sit outside that meter. Once someone becomes a client, move them to a main line so retention calls are not billed as tracking minutes.
You also do not need keyword level call tracking on day one. Campaign level attribution is enough to decide which campaigns deserve budget. Move to ad group or keyword detail only when volume and margin support the extra complexity.
How to right-size the pool without starving attribution
Step 1: Count concurrent callers, not creatives
Pull a normal peak hour for the site. Estimate how many visitors could be active at once and might call within the session window. That concurrency number is your pool floor. Add a buffer for spikes, then stop. Do not multiply by ad count.
Step 2: Start dynamic on the website, static for offline
Replace the site phone number with dynamic number insertion through your call tracking software. Keep fixed numbers for offline media. Wire website call conversions into Google Ads so the conversion action you care about is visible. Leave Call extensions and Call Only Ads on their own conversion actions until you decide which ones should be primary.
Step 3: Decide what trains Smart Bidding
Every answered ring is a soft signal. If short calls, wrong numbers, and tire kickers are primary conversion actions, Smart Bidding learns to buy more of those. Mark qualified calls or booked calls as primary when you can. Keep raw phone calls as secondary while you clean quality. Push verified outcomes back with offline conversion tracking when the CRM knows which calls turned into jobs.
Step 4: Revisit pool size when traffic jumps
Before a busy season or a big budget increase, check whether the pool still covers concurrency. Pool dilution is when two visitors share one number and attribution gets muddy. That is an accuracy problem and a cost problem. Growing the pool for real concurrency is cheap insurance. Growing it because you launched twelve new ads is not.
If you want a practical pass on pool sizing, capture gaps, and what should count as a conversion, grab the free Call Tracking Accuracy Audit Checklist. It walks through tool versus telephony checks, session and script misses, qualified call rules, click ID persistence, and the monthly reconcile so Smart Bidding is not trained on half fiction.
Granularity can wait until the money says so
People ask whether they need a number per ad group for split tests. You can A/B test landing pages and offers without a dedicated line per variant if dynamic insertion is working and UTMs or click IDs stick to the caller. The expensive failure mode is building a museum of unused numbers while the real leak is session resets, ad blockers, or a primary conversion action that rewards junk rings.
Number ownership anxiety is common too. Before you pick a vendor on price alone, ask what happens to the inventory if you leave. Portability matters more after customers have dialed that line for months.
Native Google Ads call reporting is fine for a first signal. It runs out of depth when you need recordings, multi channel source data, junk call filters, or clean client reporting. That is when CallRail, Call Tracking Metrics, or a similar stack earns the fee. Upgrade when those jobs show up, not because five starter numbers feel sparse.